A peek into the future of e-commerce with the introduction of agentic AI…

What are the market forces making AI a decisive factor for commerce today?

Gill: I think we’re at the start of a new (fourth) industrial revolution driven by AI. Like all the previous industrial revolutions, there’s going to be a ‘before and after’. What’s different about this one is the rate of change is significantly faster. Previous industrial revolutions took years, and sometimes decades. This one is taking just months.

Agentic AI: Agentic AI has really pushed this more into the mainstream, because it’s putting AI into the hands of people who previously were not able to really exploit it to the same degree before. Now, with all of these toolsets, with all of the LLMs, it’s creating the ability for people to build applications using natural language.

Digital-native consumers: There’s a whole new generation of Internet users – the digital natives – who are going to be way more comfortable in using these tools. They’re going to look at the way social media works, and they’re going to look at the way that LLMs work. And so they’re going to be very much embracing these new technologies.

Beyond hyper-personalisation: Recommendation systems – which are at the heart of our technology – are all about knowing what are the most appropriate products to show to a given user. We all have different needs, and if we search for products and get the same results, well, then that’s not really taking into account our different needs.

Generative AI and agentic AI have the ability to provide that kind of experience, like a really wonderful sales assistant who would answer all of your questions and advise you in a really intelligent and embracing way.

Now you have the possibility to give real-time feedback and interact with advertisements in a way that’s really not that easy before… It’s not just hyper-personalised, but hyper-interactive too.

Seamless customer experience: Before, your favourite retailers couldn’t serve you 24/7 because it just wasn’t economically feasible for them to have staff on duty all the time. Now they can put these tools in the hands of their customers, so that they can log on after midnight and say “I need supplies delivered tomorrow.” While that wasn’t feasible before, now you can simply do this.

And they can give you really intelligent answers with technologies like MCP and agents-to-agents: they can actually connect their agents directly to a whole series of other agents that can carry out action items in your path, on your behalf. It could be shipping or it could be returns; it could be buying complimentary products or doing reservations on your behalf.

I believe that “AI works when well done.” What constitutes “well done” with regards to AI-driven agentic commerce?

Gill: AI is “well done” when it starts from the real needs of the end user (consumer, marketer, ad‑ops) rather than what the tech can do. We don’t let the technology lead us too far. We still need to put the human at the center of everything we do.

It’s not about what we want to build. It’s what users want to use. We should never forget who we’re building it for… so it’s important to get constant feedback from them as to whether or not you’re really delivering the value you think you are.

The people who are using it at the end of the day must feel that it has been a rewarding experience to use this technology. AI is well done when the experience feels special to the end user.

For instance, hyper-personalisation before was almost a little bit in one direction, because it was difficult for the consumer to be able to give feedback to the app.

Where Criteo has a big advantage here is because we’re connected with such a very wide range of retailers and e-commerce providers across the world, and we have this huge stream of really, really rich commerce data. We are ingesting billions and billions of products per day, which gives us all of the detailed information – such as descriptions and images – about each individual product.

It’s really important for us that the technology isn’t ever used in any discriminatory fashion, and so on. That’s something that’s deep within the culture of our teams and within the culture of the company.

At Criteo, we have a Product Ethics Council, ensuring that we don’t allow our technology to show any biases. We don’t ever lose the ability to do critical thinking. You always have to see what comes back from these tools and ask yourself: ‘OK, is this really the right answer?’

And we’re always very, very keen to make sure that whatever we do, we follow only the strictest standards of privacy. Across the industry, I think it’s super important that we’re able to explain to the end users what happens with their data and how their data is used for their benefit.

How do we know what is the best combination/balance between AI ‘hitech-touch’ and human touch?

Gill: We still need to put the human at the center of everything we do. The platform can make recommendations, but you still need a human in the loop. And I think that’s really super important. It boils down to who makes that final decision. If it’s left to AI, then we are in trouble.

Giving as much information as you can, and then also providing as much choice as you can, creates the concept of informed consent… putting control in the end-users’ hands. Ultimately, I believe it’s they who should be deciding what happens and what kind of experience they have.

Bookmark